-
(making a new PR since I missed creating the design doc earlier)
RE: #1246
# IDGI API Design
## Background
IDGI (Important Direction Integrated Gradients) is a generalized framework that can b…
-
### Problem description
```toml
[project]
name = "torchaudio"
channels = ["conda-forge"]
[tasks]
[dependencies]
captum = {version="*", channel="pytorch"}
```
While trying to install this …
-
## 🐛 Bug
Between captum 0.5.0 and 0.6.0 the arguments `grad_input` and `grad_output` in the DeepLIFT implementation switched from being a tuple of length 1 to just being a PyTorch tensor. Some of t…
-
## 📚 Documentation
This is in reference to the tutorial page below.
https://captum.ai/tutorials/Llama2_LLM_Attribution
I could not find the example for LLMGradientAttribution for LLAMA2.
A…
-
![image](https://github.com/divelab/DIG/assets/117804178/2af5fdbf-26ff-4176-a47b-d15827384cb0)
-
### 🚀 The feature, motivation and pitch
Currently, we can use `to_captum` to explain one prediction example only. However, it would be good to be able to explain predictions in batches as well.
Cur…
-
Hi @rachtibat,
Great job on AttnLRP, your LRP adaptation seems very promising to attribute the behavior of Transformer-based LMs!
Provided you guys are still working on the codebase, I was wonde…
-
See #44 for more details about the dataset and network architecture.
The task is:
1. Train a baseline VQA model with decent accuracy. Try using tools from MONAI as much as possible. (related to …
-
Hey there, I've recently tested your method via the implementation in the [Inseq library](https://inseq.readthedocs.io/main_classes/feature_attribution.html#inseq.attr.feat.DiscretizedIntegratedGradie…
-
### 🛠 Proposed Refactor
The algorithms supported by captum can be easily extended in the explain module
### Suggest a potential alternative/fix
In captum_explainer.py, supported_methods=[] can be a…